I think is referring to abstractions in general. Levels and levels of abstractions. Not always a bad idea but sometimes it ends up complicating things too much and burning new comers brains.
The levels of abstraction seem to be a hazing ritual that seems to be a point of pride sometimes. I also imagine it's good job security for those who spent years on these tools.
I went from using Singularity to bundle scientific tools for HPC cluster useage at my university straight into admining OpenShift at my current employer and I have to say it was not a smooth transition. Felt like a complete idiot for a good amount of time. But after spending a few months with it I’ve really gotten to see the power and use for the k8s infrastructure and its value as a tool/platform.
And honestly, I’m digging it. It’s really cool stuff and once you get developers writing production services and applications that would otherwise have been virtualized on dedicated systems on it, they love it to for their speed of delivery and development. But we’re also managing hundreds of projects with multiple applications/services with high availability, so OCP/Kube makes sense for us.
Repeating my reply to this question asked by someone else above. I genuinely want to engage people on this question.
---
My problem with Kubernetes is people adopt it without understanding it. They adopt it because everyone else is without ever asking, "Why?"
I've installed and managed a simple K8s cluster before now and it was a nightmare (this was only a year ago.) It wasn't until we moved the management of the cluster to a managed service that it became worth it. It was at THAT point it was worth containerising the application(s) (another complexity to overcome) and letting K8s handle the deployment and scaling.
I'm worried that as an engineering culture we're not pushing back enough and saying, "Yeah K8s is cool but look: do you really need it? It's very likely you don't".
For me to automatically recommend K8s to my clients I'd need to see a (near) zero friction deployment options that's fully feature compliant and truely fully managed. It's at that point I would be happy to say, "Sure! Let's go with K8s and just write YAML files to deploy our containers!"
(And then there's the whole debate as to whether or not you even need a container around your monolith... probably not. Then the next debate about people starting out with microservices when we all know that's not a good idea.)
